invoices_from_disk_sample_code.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "gUH91JWlJAQX"
   },
   "source": [
    "# sample code: reading invoice from disk\n",
    "The generated invoices we stored in individual csv files on disk have the following format:\n",
    "```\n",
    "Pandas Dataframe with three columns:\n",
    "Invoice  Target  Truth\n",
    "                 _____\n",
    "                   |\n",
    "                   V\n",
    "   ________________________________________________________________________\n",
    "   \"uuid\", \"Sender Name\", \"Sender KvK\", \"Sender IBAN\", \"Reference\", \"Total\"\n",
    "```\n",
    "\n",
    "* \"Invoice\" contains the full text of the invoice, as generated by whatever process. It is now a string object.\n",
    "* \"Target\" is a stringified array of Jannes' target array. Its elements correspond to positions in the string \"Invoice\". Each element contains one value: a number in the range [0-5] denoting the class of each string.\n",
    "* \"Truth is a string object containing six strings, seperated by commas. \n",
    "\n",
    "Target and Truth are serialized arrays; deserialize the string back into an array with eval().\n",
    "\n",
    "The indexes in \"Truth\" correspond with those in \"Target\", except index 0. The zero-th index in \"Target\" holds the class <unknown>. In \"Truth\" it holds the uuid of the invoice.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "MMbsuHCpE1JW"
   },
   "source": [
    "# Do It Yourself\n",
    "Because google drive API will not do this for me:\n",
    "\n",
    "Download [invoices_train.zip](https://drive.google.com/file/d/1oBT6NP0y6V6xBwVdLvIB_0Sk1NjWpUUZ)\n",
    "via your local machine on to the notebook runtime environment.\n",
    "Do not try \"save link as\", just click it and visit the Google drive in a new browser tab.\n",
    "\n",
    "Make sure it is still named **invoices_train.zip** when it reaches the notebook runtime.\n",
    "\n",
    "### I tried these already\n",
    "* colab snippet for downloading a file from google drive: seems not to work for (large?) zip files\n",
    "* wget the shareable link from google drive: only downloads a silly HTML file\n",
    "\n",
    "### Alternatives?\n",
    "* Use a separate github repo for the large datasets and do a !git clone.\n",
    "* recursive wget?\n",
    "\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 204,
     "output_extras": [
      {
       "item_id": 1
      }
     ]
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 619,
     "status": "ok",
     "timestamp": 1521190653650,
     "user": {
      "displayName": "rik",
      "photoUrl": "//lh6.googleusercontent.com/-ZGIi-vN4KMU/AAAAAAAAAAI/AAAAAAAAAFk/Ki-nNio6VVM/s50-c-k-no/photo.jpg",
      "userId": "113719912521762461555"
     },
     "user_tz": -60
    },
    "id": "vPwoeWxK6D15",
    "outputId": "9f7d8f3f-05f3-4ac0-f9e9-74d6d410a9cd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Interactive namespace is empty.\n"
     ]
    }
   ],
   "source": [
    "whos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import glob\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     },
     "base_uri": "https://localhost:8080/",
     "height": 75,
     "output_extras": [
      {
       "item_id": 1
      },
      {
       "item_id": 2
      }
     ],
     "resources": {
      "http://localhost:8080/nbextensions/google.colab/files.js": {
       "data": "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",
       "headers": [
        [
         "content-type",
         "application/javascript"
        ]
       ],
       "ok": true,
       "status": 200,
       "status_text": ""
      }
     }
    },
    "colab_type": "code",
    "id": "6dFr5QsLGEvl",
    "outputId": "5fe3d3e7-dc8a-49fc-e408-c6dc74d04989"
   },
   "source": [
    "from google.colab import files\n",
    "\n",
    "uploaded = files.upload()\n",
    "\n",
    "for fn in uploaded.keys():\n",
    "  print('User uploaded file \"{name}\" with length {length} bytes'.format(\n",
    "      name=fn, length=len(uploaded[fn])))"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "eVuXv2f_BJAB"
   },
   "source": [
    "! ls *zip\n",
    "! unzip invoices_train.zip\n",
    "! ls -ld train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "autoexec": {
      "startup": false,
      "wait_interval": 0
     }
    },
    "colab_type": "code",
    "id": "5U69d52YJAQr"
   },
   "outputs": [],
   "source": [
    "train_dir = \"./train/\""
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "# ! ls {train_dir}/*.csv   # raises \"too many arguments!\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def select_batch(start=0, batch_size=32, stop=32):\n",
    "    while start < stop:\n",
    "        yield [start, min(stop, start + batch_size)]\n",
    "        start += batch_size\n",
    "#"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def select_filebatch(filenames=[], start=0, batch_size=32, total=32):\n",
    "    stop = min(len(filenames)+1, total)\n",
    "    for (first_idx, last_idx) in select_batch(start=0, batch_size=batch_size, stop=stop):\n",
    "        yield filenames[first_idx:last_idx]\n",
    "#"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def select_invoicebatch(filenames=[], batch_size=32, total=32):\n",
    "    \"\"\"yields list of invoices, list of targets, list of truths\"\"\"\n",
    "    for file_batch in select_filebatch(filenames=filenames, batch_size=batch_size, total=total):\n",
    "        invoices = []\n",
    "        targets = []\n",
    "        truths = []\n",
    "\n",
    "        for file in file_batch:\n",
    "            mysample = pd.read_csv(file)\n",
    "            # each file only contains one row, that's why we get away with the 0 in .loc[0,'invoice']\n",
    "            # else we needed to start another level of iteration\n",
    "            invoice  = mysample.loc[0,'invoice']\n",
    "            target   = eval(mysample.loc[0,'target'])\n",
    "            truth    = eval(mysample.loc[0,'truth'])\n",
    "            invoices.append(invoice)\n",
    "            targets.append(target)\n",
    "            truths.append(truth)\n",
    "        #\n",
    "        yield invoices, targets, truths\n",
    "    #\n",
    "#"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "109503 files found in directory\n"
     ]
    }
   ],
   "source": [
    "# Example of using the data and functions:\n",
    "\n",
    "filenames_all = glob.glob(train_dir + \"*.csv\")\n",
    "print(\"{} files found in directory\".format(len(filenames_all)))\n",
    "\n",
    "batch_size = 32\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### workbatch is a Python \"generator\" \n",
    "that returns something every time it is asked to behave like an iterable\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "   retrieved batch of 3 invoices\n",
      " ING Bank N.V. DATUM: 25 februari 2016 FACTUURNR. 2016130Roggeakker 16 VOOR: 3773 AB Barne\n",
      "011111111111110000000000000000000000000000000000000000000000000000000000000000000000000000\n",
      "ING Bank N.V.\n",
      "\n",
      "   retrieved batch of 3 invoices\n",
      "6-23 #IN005229  ING Bank N.V. Loberingemaat 6 7942 JD Meppel Nederland Afleveradres r de R\n",
      "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000\n",
      "ING Bank N.V.\n",
      "\n",
      "   retrieved batch of 1 invoices\n",
      "016-06-23 #IN005229  ORTEC Finance B.V. Loberingemaat 6 7942 JD Meppel Nederland Afleverad\n",
      "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000\n",
      "ORTEC Finance B.V.\n"
     ]
    }
   ],
   "source": [
    "#for instance, in a for-loop:\n",
    "workbatch = select_invoicebatch(filenames=filenames_all, batch_size=3, total=7)  \n",
    "\n",
    "piece = slice(180,270)\n",
    "for invoices, targets, truths in workbatch:\n",
    "    print()\n",
    "    print(\"   retrieved batch of {} invoices\".format(len(invoices)))\n",
    "    # print some of the retrieved content, just for the first invoice in the batch\n",
    "    print((invoices[0][piece]).replace(\"\\n\", \" \")) #print a piece of the invoice, replace newlines\n",
    "    print(\"\".join([str(x) for x in targets[0][piece]]) ) #print corresponding piece of target\n",
    "    print(truths[0][1])  # print the True Sender Name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "workbatch = select_invoicebatch(filenames=filenames_all, batch_size=16, total=30000)  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "   retrieved batch of 16 invoices\n",
      " 1of 3 1982 KB Juinen,Nederland T: 06-20180315 Betaalmethode Verzendmethode Verzendwijze -\n",
      "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000\n",
      "Unilever Nederland\n"
     ]
    }
   ],
   "source": [
    "# for instance using the next() function\n",
    "# hit CTRL-Enter for multiple retrievals\n",
    "\n",
    "invoices, targets, truths = next(workbatch)\n",
    "\n",
    "print()\n",
    "print(\"   retrieved batch of {} invoices\".format(len(invoices)))\n",
    "# print some of the retrieved content, just for the first invoice in the batch\n",
    "print((invoices[0][piece]).replace(\"\\n\", \" \")) #print a piece of the invoice, replace newlines\n",
    "print(\"\".join([str(x) for x in targets[0][piece]]) ) #print corresponding piece of target\n",
    "print(truths[0][1])  # print the True Sender Name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "# it is even possible (I hope?) to use the object directly as an input for a keras model\n",
    "\n",
    "# although, models need a lot of data-preprocessing before you get to training/evaluating/predicting\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "default_view": {},
   "name": "invoice_from_disk_sample_code.ipynb",
   "provenance": [
    {
     "file_id": "1wjtiXpxRkuN5V57GiK_RSPxy84cFbERO",
     "timestamp": 1520877467293
    }
   ],
   "version": "0.3.2",
   "views": {}
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}